Lakera

Lakera Guard screens LLM prompts and responses for prompt injection, policy violations, and PII.

Lakera Guard screens prompts and responses via the /v2/guard API — blocking on policy hits and redacting PII spans when only pii/* detectors fire.

Requires Backend v1.16.2+.

Using Lakera with Portkey

1. Add Lakera Credentials to Portkey

• Navigate to Admin Settings → Plugins
• Edit the Lakera integration
• Add your Lakera API Key (create a Guard API key at platform.lakera.ai)
• Optionally set API Base URL (defaults to https://api.lakera.ai)

2. Add Lakera Guardrail Check

• Navigate to Guardrails → Create
• Search for “Lakera Guard” and click Add
• Configure:
  • Lakera Project ID (projectID, optional) — project whose policy defines detectors to run
  • Payload End Offset Is Inclusive (endInclusive, default false)
  • Timeout (timeout, default 30000 ms)
• Set guardrail actions and create the guardrail

Check Name	Description	Parameters	Supported Hooks
Lakera Guard	Calls Lakera /v2/guard with payload+breakdown	projectID, endInclusive, timeout	beforeRequestHook, afterRequestHook

3. Attach to a Config

{
  "input_guardrails": ["guardrails-id-xxx"],
  "output_guardrails": ["guardrails-id-yyy"]
}

See Configs for usage with SDKs and the gateway.